We recovered the IR injection alignment. The IR injection power was 430uW. The mode matching is 94.3%. Note that the misalignment is combination of pitch and yaw (top right and bottom left sides are bright).

I brought in PCI the cleaned viewport. Aso-san helped me confirm that the cleaned surface is the one where there is no more indium at the edge.

I installed this viewport on the holder with the correct surface facing the lasers.

I realigned DC and started the measurement.

The procedure I'm planning to do is to start from 28 mW and increase step by step up to 600 mW (eg 100,200,30, etc..)

I did the first measurement with 28 mW and got the result attached in figure 1.

It seems that we don't have clear signal on most of the map. But there are few points where the absorption seems quite high.

Sadly I was not present in the PCI during this measurement so I can not confirm that there was no saturation of the lockin so these maximal values of absorption are not reliable.

A possibility for further inspection is to repeat this measurement with same power but only around the dirty spots and make sure that the lockin is not saturating.

Depending on the absorption value of these spots, include them or not in the maps with higher laser power.

I got a bit sceptical of the spare viewport measurement reported in entry 2580.

Indeed the absorption seems too uniform with really low value..

• Discussing with Matteo, a possibility could be that the spare viewport is tilted (indeed it is placed inside a tama size holder with the addition of 2 spacers to compensate the 1 cm thickness of the viewport). That would make the surface calibration not reliable.

So I reinstalled spare viewport and aligned the DC with Pin = 0.6248W.

First, I checked the tilt of the viewport by doing Z scan and checking the the Z center of the absorption (computed from the 2 positions where the phase is 0 deg) .

From these measurements, it seems that the difference of the Z center over the radius of the absorption map is ~0.35 mm. This is far smaller than the interaction length ~2 mm (see Manuel PhD).

So we can conclude that there is negligeable tilt with this holder.

• Another possibility could be that the lockin sensitivity was saturating at a low value, making a constant noise level fakes an absorption signal.

However, an easy check is to double the power and check if the AC signal is also doubled. That would confirm that the signal we are seeing is coming from absorption.

Therefore, I decided to do a new measurement with Pin = 1.2545 W (radius 30mm and step size 2 mm) which is roughly twice larger than the measurement reported in entry 2580.

The mean ac value went from 8.1 e-4 to 1.6e-3. This confirms that we are seeing absorption signal !

Finally I also took a new absorption map with this higher power as reported in figure 1.

In figure 2, you can see the same map but with the colorscale constraint between 15 and 20 ppm and we can see the (small)  non-uniformity of the viewport absorption.

I fitted the power as a function of the HWP angle.

Quite useful to know how to increase the power step by step.

[Aritomi, Yuhang, Michael, Marc]

Today the new mirror mounts (FMD MM1000S) arrived and we started the replacement work.

We replaced 1 mirror mount before AMC and 3 mirror mounts in IR injection path.

We also replaced 5 mirrors in the IR injection path with layertec mirror (106910) whose reflectivity is more than 99.9%.

While I mounted the mirror, I tightened the screw too much and broke one mirror...

After the replacement work, we completely lost the IR alignment, but we recovered it. Now the IR transmission is more than 400 with 450uW injection. The IR transmission should be more than 500 with 450uW injection, so we still need to optimize it.

From tomorrow, we will optimize the injection alignment and replace mirrors and mirror mounts in the reflection path and try to optimize the reflection mode matching.

[Aritomi, Michael]

We measured squeezing and anti-squeezing for different green power with CCFC. We changed the green power from 10mW to 30mW with 5mW step. The OPO temperature and p pol PLL frequency are summarized in elog2577.

For green power larger than 40mW, we couldn't find squeezing (~ -130dBm at minimum) and the noise level suddenly changed from -130dBm to -110dBm...

The result is shown in the attached figure. Since we couldn't inject large green power, the phase noise information cannot be extracted.

I measured the round trip losses with new beam spot which has more stable IR transmission. The round trip losses are 139 ppm.

P_res = 245; % BAB reflection power on resonance (uW)

P_in = 362;   % BAB reflection power off resonance (uW)

R_gamma = P_res/P_in;

gamma = 0.05; % uncoupled power to the cavity: 5% from mode mismatch

R = (R_gamma-gamma)/(1-gamma);

T = 0.00136; % input mirror transmissivity

L = (T/2)*(1-R)/(1+R) = 139 ppm

Despite 2 windows updates that turned off the computer I finally finished the measurement of the absorption map of the spare viewport (see fig 1).

In order to recover the lost time I changed the step size to 0.5 mm.

The mean absorption is at the order of 18 ppm while there is some dust/surface defects creating few spikes in the absorption.

Marc and Yuhang

We ran mcmc fit of measurement in elog2546 based on the emcee pakage. This run took about 3 hours by using a computer with 20 cores, which is ten times faster than the time of 30 hours by using a single core. This time consuming is reasonable since there will be some overhead while using multiprocessing.

In this run, we tried two situations: two free parameters (detuning, homodyne angle), four free parameters (squeezing level, optical losses, detuning, homodyne angle).

The mcmc run for two free parameters is shown in attached figure 1 and 2. This run is very successful since we see a clear contour distribution in figure 1 and parameters are always in the stable region as figure 2.

The mcmc run for four free parameters is shown in attached figure 3 and 4. This run needs some optimization.

I measured the IR injection mode matching with new beam spot which has more stable IR transmission.

First I optimized IR injection alignment with new beam spot. I checked the IR injection alignment by changing AOM frequency. The mode matching is 95.6%. Note that AA and BS pointing were engaged during this measurement. 

After the IR injection alignment, IR transmission was 460 with 430uW injection. According to our experience, the IR transmission should be 490 with 430uW injection. The possible reason of the smaller IR transmission is that the IR transmission is off centered in transmission PD with new beam spot.

I recorded the thermistor (temperature) values that I used for the OPO nonlinear gain measurement.

Initially I searched in increments of +/- 0.01 kOhm on the thermistor, and then checked a bit within the optimal range. Perhaps there is further room for optimisation when zooming in on the oscilloscope though. At 100 mV/div ranges on the oscilloscope, the difference in voltage for thermistor change under 0.01 kOhm was hard to distinguish.

In the table I give the temperature, oscilloscope reading as well as the range value on the power meter - using a range of 1.6 mW on the power meter gives 10x more voltage on the oscilloscope than a range of 16 mW.

Today I increased the laser diode current to 3A.

I also took compared the output power as a function of HWP angle :

I set the HWP to 49 degrees did a large z scan and could see some AC signal.

I checked the viewport center :

Z_center = 73.4 mm (from phase = 0)

Y_center = 120.95 mm (from ac/dc = 0)

X_center ~ 331.75 (one of the extrema was not possible to estimate because I placed a stop in Zaber to avoid burning the holder.

The X_center was estimated from the top X value of 383.9 mm and assuming a similar radius as the Y direction (52.15 mm)

I tuned the DC at this position and got as a preliminary absorption value :

corr = 1;
ac = 8e-4;
dc = 4.08;
Pin = 0.613;
R_surf = 17.033
absorption = ac/(dc*P_in*R_surf)*corr ~416 ppm 18 ppm (Edit : I took wrong power in previous estimation...)

The absorption map is now on-going with a 3 cm radius and 50 um step size.

This measurement was done on June 10th. 

Recently the bump at 100Hz came back as shown in an attached figure.

Marc and Yuhang

Marc and I were suspecting that the beam may not hit on the center of mirror. To check that, we decide to scan BS pitch/yaw and check IR transmission.

Today I did this check. I found two resonant beams while scanning BS, which can be seen from the attached video (two videos show the situation of pitch and yaw separately).

Yaw: https://drive.google.com/file/d/1ahDx6MTfYqxdoFXbyH0PbrLRPybCtIhp/view?usp=sharing

Pitch: https://drive.google.com/file/d/1N6N3BdUr3WIG_GeNLJeDQDmCbeHzdTXm/view?usp=sharing

By pointing the incident beam towards the center of mirror, these two resonant beams get closer. In the end, they still don't overlap and separate by roughly a HWHM. In the end, the pointing to the center has a precision of 2urad, considering the distance of 300m, this is 0.6mm precision.

After centering the beam, I measured the IR length noise by using BAB. Attached figure 1 shows their difference. We can see that the low frequency length change is greatly reduced. Note: the calibration of IR length is done by taking the slope of PDH signal as attached figure 2, 3, 4, 5 for on resonance and detuned case, which may introduce some error. Especially, the calibration of the green line (beam off-center and on resonance) was using a PDH signal with opposite slope compared with others.

The attached figure 6 shows the new location of beam now. This moves the beam by 6mm in both yaw and pitch on end mirror. We should go to end room and center beam on both camera and PSD.

In addition, the AOM frequency changed by 2415Hz, which indicates the GR and IR length difference is 2415/2e-12m = 1.2075nm. This tells us, with the residual pointing error of IR, we have 1nm GR/IR length difference after beam moves 6mm on mirror. This phenomon explains the IR detuning change issue for different alignment of GR.

This entry summarizes the (too many) measurements performed last Thursday and Friday...

I started by checking the bulk reference sample as it was the last calibration measurement performed ( see entry 2510).

First, I had to change the HWP that controls the ir laser power from 118 degrees to 119 degrees in order to get the usual Pin~28 mW.

Then, I found out that the calibration was lower than expected at 0.5 cm/W.

This, together with the (small) laser power change seems to indicate that the beam parameters changed during a 2 weeks window.

My guess is that, the temperature/humidity increase of last week affected the laser beam parameters (effects of temperature inside TAMA central building can be found in entry 2493 where PR pitch motion seems to follow trends similar to temperature change troughout the day).

I replaced the bulk reference sample by the surface reference sample and started several measurements...

To summarize :

I checked the IU optimal position : 68 mm is still good.

I checked the sample Z position : after several tests, I found out that a good position for the surface reference sample is at Z = 39.5 mm

This corresponds indeed to have the surface of this sample at the ir beam waist and crossing with the red beam (details in entry 2446)

With the good sample Z position, I still could not achieve 'optimal' calibration factor.

I checked again the IU position and confirmed than IU at 68 mm is optimal.

I found out that having this sample Y center (ie Y = 121.2 mm) at the beam crossing was giving :

y = 121.2

AC_surfref = 0.4075;

DC_surfref = 3.966;

acdc = 0.1026;

P_in = 27.6e-3;

R_surf = 16.8972

while moving the Y position to 120.2 mm gives

AC_surfref = 0.411;

DC_surfref = 3.96;

acdc = 0.1038;

P_in = 27.6e-3;

R_surf = 17.0949

I will need to check if this sample has some dust in the center...

Finally, I moved on the spare viewport measurement :

I placed it on the TAMA holder using 2 spacers to fix it.

The good Z position to have the surface at the 2 beams crossing point is z = 38 + 60/2 = 68 mm

The good IU Z position is 68 - 3.4 = 64.6 mm (here I assumed 1 cm thickness of the viewport).

Unfortunately, at that point I moved by mistake the last lens on the IR path... So I performed again the surface calibration as :

AC_surfref = 0.4125;

DC_surfref = 3.974;

acdc = 0.1038;

P_in = 27.7e-3;

abs_surfref = 0.22;

R_surf = 17.0331 (screenshot saved at 20h35)

I re-installed the spare viewport and tried to measure its absorption.

28 mW did not show any signal.

428 mW (ie HWP at 89 degrees)  did not show anything as well...

As this power is around the maximal that can be produced with the laser diode current of 1 A and because any current change requires ~1h to stabilize the laser power, I stopped the measurement at that point.

Anyway, this tells us that the absorption is quite low...

Good measurement! Now it is obvious that the OPO threshold is lower than before as I said. 

Retry of the OPO nonlinear gain measurement in 2528, motivated by the idea that the uncertainty in the detuning fit (i.e. 2544, 2546, 2550) may be caused by uncertainty in the fit parameters (see 2544, 2549).

The previous measurement of OPO nonlinear gain was too imprecise in a way that user error had a lot of impact. In the previous process, we left the OPO unlocked while sending a sine wave to GRPS to modulate between amplification and deamplification. We then used the oscilloscope persist function and triggered the resonant peak to appear on the screen. The way this was intended to work is shown in figure 1, however, it seemed like the peak was fluctuating in the horizontal axis too much due to the cavity unlock, and obtaining the minimum value was quite imprecise. On top of that, the OPO temperature was not optimised for different green injection power to compensate for absorption. This of course causes an error in the co-resonance of p-pol/BAB and green.

This time I measured again with the OPO locked, and optimised the temperature and PLL frequency for each value of green injection power. With the cavity locked, the output transmission is just a sine wave due to GRPS modulation, going between amplification and deamplification. While there is less uncertainty in just reading off the values now, there is of course some user uncertainty in optimising the temperature and PLL frequency, especially because being too careful about it is very time consuming. Juding by the optimisation process, I would say that the temperature is easily optimised to about +/- 0.01 kOhm on the thermistor, with differences in this range amounting to about +/- 2-8 mV on the amplification reading on the oscilloscope, for values between 100-1000 mV.

The values were measured using a power meter in transmission of the OPO. There are a lot of green junk beams coming out of the OPO as well, so they were removed with a laser line filter to leave only the BAB transmission. I measured a calibration factor of 0.00451 mW/mV for the correspondance between power meter readout and oscilloscope voltage, but in retrospect this is unnecessary, since only the relative power matters, and the oscilloscope voltage is linear with power meter incident optical power. Aritomi also has a more accurate calibration in 2566.

Figure 2 shows the OPO nonlinear gain with simple curve fitting. I will further consider the error of the input parameters in a future update.

I analyzed again with correct shot noise level (-133.7dBVrms/rtHz) and 12% loss from visibility. The readout loss includes 12% loss from visibility, 2% loss from homodyne (dark noise, quantum efficiency, AR), and 20% loss from pick off BS.

sqz_dB = 11.8;                  % produced SQZ (dB)

L_rt = 120e-6;                  % FC losses

L_inj = 0.32;                   % Injection losses

L_ro = 0.31;                    % Readout losses

A0 = 0.05;                      % Squeezer/filter cavity mode mismatch

C0 = 0.06;                      % Squeezer/local oscillator mode mismatch

ERR_L =   1e-12;                % Lock accuracy (m)

ERR_csi = 30e-3;                % Phase noise (rad)

[Aritomi, Michael]

We measured the visibility between LO and BAB reflected from FC for 10s. The BAB was half detuned.

To know the BAB power during the measurement, we also measured the BAB pick off power in CCFC port. Fig 1 shows the BAB power before AMC and the BAB pick off power. From this measurement, we can know the calibration factor from the BAB pick off power to the BAB power before AMC (bottom plot in Fig 1). The calibration factor is 0.246. The LO power was 1.16V. The offset of visibility was 9mV and the offset of pick off was 5.8mV. Note that the offset subtraction is important for visibility calculation.

The measured visibility and BAB pick off are shown in Fig 2. The 10Hz modulation was applied to IR phase shifter. As you can see, both of visibility and BAB pick off are fluctuating. To calculate the visibility, we divided this data into 1000 segments with 0.01s step (Fig 3) and calculated the visibility in each segment (Fig 4). From the histogram, the visibility is 0.94(1) which corresponds to optical loss of 12(2)%.

In FDS measurement in PRL paper, we assumed that the visibility is 0.98 which corresponds to optical loss of 4%. It seems that the visibility got worse and we have 8% more loss than before...

I measured BAB power fluctuation after OPO for one minute (attached picture). The BAB power fluctuation is 0.7% (Ratio of Max/Min = 1.007).